Asynchronous canvas drawing with Web Workers - javascript

So, I'm making a relatively trivial HTML5 Canvas drawing web app. Basically, you can select your color, and then draw on a 500x500 canvas. It's going to be themed as a "graffiti" wall, so I am attempting to create a graffiti effect to the drawing, much like the Spray tool in MS Paint of yore.
Feel free to take a look at it here.
In order to facilitate this effect, I'm making use of web workers to callback on mouse events and asynchronously draw to the canvas. The naive implementation I have now is that on any mouse event, 5 pixels are randomly drawn around the coords of the event.
What I would like to do though, is to have those pixels drawn continuously from the mousedown event until the mouseup event, while updating the coords on mousemove events. From my limited knowledge of JavaScript, I imagine that this could involve a setTimeout(), but I'm not sure how to manipulate this to achieve what I want.
DISCLAIMER: This is part of a school project, and as such I am trying to avoid JQuery, Ajax, and other such frameworks; my goal here is to make an as-pure-as-possible JavaScript/HTML5 web app.
Thanks in advance.

Using a timer (no worker required):
var mouseX = 0,
mouseY = 0,
mouseDown = false;
function ev_canvas( ev ) {
if (ev.offsetX || ev.offsetX == 0) { //opera
mouseX = ev.offsetX;
mouxeY = ev.offsetY;
} else if (ev.layerX || ev.layerX == 0) { //firefox
var canvasOffset = document.getElementById("graffiti_wall").getBoundingClientRect();
mouseX = ev.layerX - canvasOffset.left;
mouseY = ev.layerY - canvasOffset.top;
}
if ( ev.type == 'mousedown' ) {
mouseDown = true;
}
else if ( ev.type == 'mouseup' ) {
mouseDown = false;
}
}
function draw_spray() {
if( !mouseDown ) {
//Don't do anything since the mouse is not pressed down
return;
}
//Draw something at the last known location
context.strokeRect( mouseX, mouseY, 1, 1 );
}
//Call draw_spray function continuously every 16 milliseconds
window.setInterval( draw_spray, 16 );

If you want to use a WebWorker (for example for more complex drawing algorithms), I could think of the following setup:
onmousedown, spawn a new worker and register a handler on it that paints objects on the canvas
onmouseup (and -leave etc), terminate the worker
onmousemove, if a worker exists determine mouse coordinates and send them into the worker
In the worker
listen to new mouse coordinates
start an timeout interval which constantly fires draw-events (depending on current coordinates and a clever algorithm)
Yet I think that a Worker is too much overhead for a simple graffiti tool. Use the simple solution without a Worker like #Esailija demonstrated.
If you had a more complex application which could make good use of Workers, you wouldn't really spawn them onmousedown and terminate them. Instead, you maybe instantiated single Workers for single kinds of tools, and fired start-processing and end-processing events to them.

Related

CreateJS click target is wrong on Samsung Internet browser

I develop a html5 canvas browser game with createjs (Adobe Animate) and after the last update of Samsung Internet browser when the user clicks somewhere, it fires the click event for totally different button. It looks like the createjs native method _getObjectsUnderPoint() gives the wrong target of the click. Even when I click on empty space on the stage, where there is not buttons, a click handler is fired. This strange behavior is not present in Google Chrome or Opera for mobile, only in Samsung Internet.
In rare cases, even though the click target is correct, the handler is not called. More specifically, the problem must be in the createjs functions _testMask() or _testHit(). They use the canvas context and matrix transformations.
Does anyone has idea what should be fixed in the CreateJs library for Samsung Internet browser.
I solved the problem like this:
Check if the user agent contains SamsungBrowser
var userAgent = navigator.userAgent;
userAgent = userAgent.toLowerCase();
var isSamsungBrowser = userAgent.indexOf("samsungbrowser") != -1;
In CreateJs in _getObjectsUnderPoint() I added this near the end of the
for loop:
if(isSamsungBrowser) {
var bounds = child.getBounds();
if(bounds) {
var p = child.globalToLocal(x, y);
var hit = p.x >= bounds.x &&
p.x <= bounds.x + bounds.width &&
p.y >= bounds.y &&
p.y <= bounds.y + bounds.height;
if(! hit) {
continue;
}
}
else {
console.error('NO BOUNDS FOR', child);
continue;
}
}
The only drawback is that every Shape must have bounds. So you can convert the shapes to bitmaps in Adobe Animate or use setBounds().

Queue getting multiple video frames

I'm creating a PDF output tool using jsPDF but need to add multiple pages, each holding a canvas image of a video frame.
I am stuck on the logic as to the best way to achieve this as I can't reconcile how to queue the operations and wait on events to achieve the best result.
To start I have a video loaded into a video tag and can get or set its seek point simply with:
video.currentTime
I also have an array of video seconds like the following:
var vidSecs = [1,9,13,25,63];
What I need to do is loop through this array, seek in the video to the seconds defined in the array, create a canvas at these seconds and then add each canvas to a PDF page.
I have a create canvas from video frame function as follows:
function capture_frame(video_ctrl, width, height){
if(width == null){
width = video_ctrl.videoWidth;
}
if(height == null){
height = video_ctrl.videoHeight;
}
canvas = document.createElement('canvas');
canvas.width = width;
canvas.height = height;
var ctx = canvas.getContext('2d');
ctx.drawImage(video_ctrl, 0, 0, width, height);
return canvas;
}
This function works fine in conjunction with the following to add an image to the PDF:
function addPdfImage(pdfObj, videoObj){
pdfObj.addPage();
pdfObj.text("Image at time point X:", 10, 20);
var vidImage = capture_frame(videoObj, null, null);
var dataURLWidth = 0;
var dataURLHeight = 0;
if(videoObj.videoWidth > pdfObj.internal.pageSize.width){
dataURLWidth = pdfObj.internal.pageSize.width;
dataURLHeight = (pdfObj.internal.pageSize.width/videoObj.videoWidth) * videoObj.videoHeight;
}else{
dataURLWidth = videoObj.videoWidth;
dataURLHeight = videoObj.videoHeight;
}
pdfObj.addImage(vidImage.toDataURL('image/jpg'), 'JPEG', 10, 50, dataURLWidth, dataURLHeight);
}
My logic confusion is how best to call these bits of code while looping through the vidSecs array as the problem is that setting the video.currentTime needs the loop to wait for the video.onseeked event to fire before code to capture the frame and add it to the PDF can be run.
I've tried the following but only get the last image as the loop has completed before the onseeked event fires and calls the frame capture code.
for(var i = 0; i < vidSecs.length; i++){
video.currentTime = vidSecs[i];
video.onseeked = function() {
addPdfImage(jsPDF_Variable, video);
};
}
Any thoughts much appreciated.
This is not a real answer but a comment, since I develop alike application and got no solution.
I am trying to extract viddeo frames from webcam live video stream and save as canvas/context, updated every 1 - 5 sec.
How to loop HTML5 webcam video + snap photo with delay and photo refresh?
I have created 2 canvases to be populated by setTimeout (5000) event and on test run I don't get 5 sec delay between canvas/contextes, sometimes, 2 5 sec. delayed contextes get populated with image at the same time.
So I am trying to implement
Draw HTML5 Video onto Canvas - Google Chrome Crash, Aw Snap
var toggle = true;
function loop() {
toggle = !toggle;
if (toggle) {
if (!v.paused) requestAnimationFrame(loop);
return;
}
/// draw video frame every 1/30 frame
ctx.drawImage(v, 0, 0);
/// loop if video is playing
if (!v.paused) requestAnimationFrame(loop);
}
to replace setInterval/setTimeout to get video and video frames properly synced
"
Use requestAnimationFrame (rAF) instead. The 20ms makes no sense. Most video runs at 30 FPS in the US (NTSC system) and at 25 FPS in Europe (PAL system). That would be 33.3ms and 40ms respectively.
"
I am afraid HTML5 provided no quality support for synced real time live video processing via canvas/ context, since HTML5 offers no proper timing since was intended to be event/s controlled and not run as real time run app code ( C, C++ ...).
My 100+ queries via search engine resulted in not a single HTML5 app I intend to develop.
What worked for me was Snap Photo from webcam video input, Click button event controlled.
If I am wrong, please correct me.
Two approaches:
create a new video element for every seek event, code provided by Chris West
reuse the video element via async/await, code provided by Adrian Wong

Multitouch Pinch, Pan, Zoom in HTML 5 Canvas

I have most of a module written to handle multitouch pinch, pan and zoom on an HTML 5 canvas element. I will share it below. I've been developing in JavaScript for some time now, and this one continues to boggle me. If anybody has any insights, I will post the final version up on stack for everybody to share once it is confirmed working on my iPad.
Here is what I am doing:
touchmove events trigger the changing of variables. I use these variables to change the way in which my image is painted onto the canvas. I have eight variables, each corresponding to the options that can be put into the drawImage() function. These eight variables get updated through functions that increment/decrement their values and keep them within a certain range. The variables are closure variables, so they are global throughout my module. To prevent over-processing, I make a call to this drawImage() function once every 40ms while the user has their finger pressed to the screen using a setInterval().
Here is the problem:
touchmove events seem to be causing a race condition where my variables get updated by many different instances of that same event. I can somewhat confirm this through my console output, that tracks one variable that is bounded to never reach below 20. When I swipe in one direction quickly, that variable dips down far below 20. Then when I release my finger, swipe slowly, it returns to 20. Another thing that points me in this direction, when I look at these variables while stepping through my program, they differ from what my console.log() pumps out.
Note: The code successfully draws the image the first time, but not anytime thereafter. A basic rendition of my code is below... The full version is on GitHub inside the Scripts folder. It is a Sencha Touch v1.1 app at heart
function PinchPanZoomFile(config)
{
/*
* Closure variable declaration here...
* Canvas Declaration here...
*/
function handleTouchStart(e) {
whatDown.oneDown = (e.originalEvent.targetTouches.length == 1) ? true : false;
whatDown.twoDown = (e.originalEvent.targetTouches.length >= 2) ? true : false;
drawInterval = setInterval(draw, 100);
}
function handleTouchEnd(e) {
whatDown.oneDown = (e.originalEvent.targetTouches.length == 1) ? true : false;
whatDown.twoDown = (e.originalEvent.targetTouches.length >= 2) ? true : false;
clearInterval(drawInterval);
}
function handleTouchMove(e) {
if(whatDown.twoDown) {
/*
* Do Panning & Zooming
*/
changeWindowXBy(deltaDistance); //deltaDistance
changeWindowYBy(deltaDistance); //deltaDistance
changeCanvasXBy(deltaX); //Pan
changeCanvasYBy(deltaY); //Pan
changeWindowDimsBy(deltaDistance*-1,deltaDistance*-1); //(deltaDistance)*-1 -- get smaller when zooming in.
changeCanvasWindowDimsBy(deltaDistance,deltaDistance); //deltaDistance -- get bigger when zooming in
} else if(whatDown.oneDown) {
/*
* Do Panning
*/
changeWindowXBy(0);
changeWindowYBy(0);
changeCanvasXBy(deltaX);
changeCanvasYBy(deltaY);
changeWindowDimsBy(0,0);
changeCanvasWindowDimsBy(0,0);
}
}
function draw() {
//Draw Image Off Screen
var offScreenCtx = offScreenCanvas[0].getContext('2d');
offScreenCtx.save();
offScreenCtx.clearRect(0, 0, canvasWidth, canvasHeight);
offScreenCtx.restore();
offScreenCtx.drawImage(base64Image,
parseInt(windowX),
parseInt(windowY),
parseInt(windowWidth),
parseInt(windowHeight),
parseInt(canvasX),
parseInt(canvasY),
parseInt(canvasWindowWidth),
parseInt(canvasWindowHeight)
);
//Draw Image On Screen
var offScreenImageData = offScreenCtx.getImageData(0, 0, canvasWidth, canvasHeight);
var onScreenCtx = canvas[0].getContext('2d');
onScreenCtx.putImageData(offScreenImageData, 0, 0);
}
}
I strongly recommend using Sencha Touch 2.0.1, as it supports a lot of the touch events you need. See http://dev.sencha.com/deploy/touch/examples/production/kitchensink/#demo/touchevents for examples.

KeyDown/KeyUp 70-120ms delay. How to reduce?

We're developing arcade (a lot of action and speed) browser 2d-game using canvas.
Sometimes our testing players report us that there is a delay: player still moving 5-10 pixels away after keyup.
I've digged this issue, you can see yourself delay http://jsfiddle.net/C4ev3/7/ (try keydown/up any key as fast as you can). My results is from 70 to 120ms. And i think that's a lot. (FYI, our network latency is 10-20ms).
Any ideas how to reduce this delay?
upd i've noticed that on good hardware this delay is under 30-40ms. But i'm testing on core2duo, winxp, chrome 19 - it's not a P4 with IE6 :)
Hi one thing you could do is instead of using an anonymous function try using defined functions,
http://jsfiddle.net/C4ev3/10/ - for me this reported at 50-100 MS
However i would not recommend jQuery for Canvas Applications it's very big for the very little you using, you should try using native Javascript
http://jsfiddle.net/C4ev3/11/ - for me this reported 30-70 MS
Javascript Threading
One thing i noticed in the comments Javascript is not Multi-Threaded Well Urm-Arr,
it sort of is setInterval is Async not Sync, however affecting the window is a single thread E.G if you have a Class that has some number is it using a setInteval will use another thread and not have a problem altering the math however in the Task then requires a Draw on the page it will enter the bottom of the JS handle Que,
Certain parts of Javascript are on a different thread how ever any thing changing the page has to run on the Main Thread same as any Windows application if your thread want to change the Form your have to invoke the main thread to do it for you
however it is not multi-threaded like any thing else you cant just handle or abort at a given Wim like windows,
Other ASync Tasks include AJAX has the option to be both Async and Sync
Updated to show my comment about FPS limiting:
Please bear with me. This is linking to a project that is allready built to show the example:
so my Game is Completely OOP
var elem = document.getElementById('myCanvas');
var context = elem.getContext("2d");
context.fillStyle = '#888';
context.lineWidth = 4;
// Draw some rectangles.
context.fillRect(0, 0, 800, 600);
context.fillStyle = '#f00';
var ball = new Ball();
var leftPadel = new Padel(10, 60, 40, 120);
var rightPadel = new Padel(750, 520, 40, 120);
pong = new Pong();
pong.draw();
setTimeout("ball.move()", pong.redrawTime());
Inside my pong class is where all the main workings of the game goes but here are the FPS bit you need to see
this.fps = 30;
this.maxFPS = 60;
this.redrawTime = function(){
return (1000 / this.fps)
}
this.lastDraw = (new Date)*1 - 1;
Then as you can see my Interval is on ball.move this calls the main pong class again on redraw at the End of the redraw i have the FPS checking and limiting code
this.fps = ((now=new Date) - this.lastDraw);
if(this.fps > this.maxFPS){
this.fps = this.maxFPS;
}
this.lastDraw = (new Date)*1 - 1;
if(this.reporting = true){
console.clear();
console.log("FPS: "+this.fps.toFixed(1))
}
setTimeout("ball.move()", pong.redrawTime());
This then forces you to get the Best Possible FPS without queuing the Main Thread
Try this:
e.stopPropagation()
Stops the bubbling of an event to parent elements, preventing any
parent handlers from being notified of the event.
e.preventDefault()
Prevents the browser from executing the default action. Use the method
isDefaultPrevented to know whether this method was ever called (on
that event object).
My min. results in Google chrome: 7ms

Javascript events responsiveness over a <canvas> element

I am writing an html5 app which manipulates video on a canvas.
I am also showing a custom (self-drawn) mouse cursor over it.
To update my app I am calling the setInterval function to draw stuff to the canvas.
As I am manipulating video I need: Very high refresh rates + Lot's of processor time to draw. (~50ms per frame).
My draw function causes my on-mouse function to starve. (this is acceptable by me).
But... After it is finished starving it responds to old events. It can take up to 3 frames for the mouse to catch so that I can render it in the right position. Meaning you can see the cursor "crawling" after you've stopped moving the mouse.
Is there a way to give the onmousemove events a higher priority then my setInterval(drawFunction)?
When in the draw function, Can I "ask" if there are pending mouse events, and revoke my current call to draw?
Is there some-other hack I can use? (I can draw to back-buffer in a webWorker, but as I understand from reading up on html5 this is only thread abstraction [threads are not concurrent] )
You can't prioritize event handling, at least not directly.
What you might consider would be having your own timer-driven code check for pending mouse events. The mouse event handler would just put a work request into a queue. Your video manipulating could could just check that queue and handle operations as it sees fit. Thus, the real mouse work would also be done in the timer code.
edit Here's what I'm thinking. You'd still have handlers for your mouse events:
var eventQueue = [];
canvasElement.onmousemove = function(evt) {
evt = evt || event;
eventQueue.push({ type: evt.type, x: evt.clientX, y: evt.clientY, when: new Date() });
};
Thus the mouse handlers would not do any actual work. They just record the details in a list of events and then return.
The main, timer-driven loop can then just check the queue of events and decide what to do. For example if it sees a whole string of "mousemove" events, it could compute the net change in position over all of them and update the cursor position just once.
You certainly can use a Web Worker to manipulate your pixels in the background. Web Workers do run concurrently. It doesn't seem to be part of the spec, but every implementation runs workers in a separate process. So your main script will be responsible for updating your custom cursor, but you'd pass the canvas ImageData off to the background worker for the video processing. Note that you can't send an actual Canvas or Context to the worker, as workers have no DOM access.
There is no way to give events higher priority, they seem to be serviced on a first come, first serve basic. Assuming you are using setTimeout, one thing that might help though is to break up your tasks into smaller restartable chunks, e.g. if you are processing an image like this:
for(y=0; y<height; y++) {
// Deal with rows of pixels
}
// Show image
You could do this instead:
var INTERVAL = height/4;
for(y = old_y; y<height && y<old_y+INTERVAL ; y++) {
// Deal with row of pixels
}
if (y == height) {
// show image
}
old_y = (y == height) ? 0 : y;
Then the other events will have 4x (or whatever, depending on the INTERVAL) greater chance of being dealt with.

Categories